mikav-ai/kerala-corpus

Kerala Corpus

All-in-one open dataset scaffold for Kerala-focused Malayalam language model training.

Structure

kerala-corpus/
  data/
    districts/
    geography/
    culture/
    history/
    language/
    government/
    economy/
    education/
    health/
    transport/
    environment/
  metadata/
  docs/
  scripts/

Record Format

Each .jsonl line should be one JSON object:

{"id":"culture-festivals-000001","domain":"culture","subdomain":"festivals","district":"thrissur","text":"...","language":"ml","source":"...","source_url":"...","license":"cc-by-4.0"}

Use only open, redistributable data and keep source/license metadata with every record.

Validate

python scripts/validate_jsonl.py